vc 13/14 t18 visual feature extraction - dcc · vc 13/14 - t18 - visual feature extraction how to...
Post on 25-Jun-2020
27 Views
Preview:
TRANSCRIPT
Mestrado em Ciência de Computadores
Mestrado Integrado em Engenharia de Redes e
Sistemas Informáticos
VC 13/14 – T18
Visual Feature Extraction
Miguel Tavares Coimbra
VC 13/14 - T18 - Visual Feature Extraction
Outline
• Feature Vectors
• Colour
• Texture
• Shape
VC 13/14 - T18 - Visual Feature Extraction
Topic: Feature Vectors
• Feature Vectors
• Colour
• Texture
• Shape
VC 13/14 - T18 - Visual Feature Extraction The earth is round
The earth is blue,
white and
brown
Visual Features
The south pole has
a smooth texture
VC 13/14 - T18 - Visual Feature Extraction
Visual Features
• Features
– Measure specific characteristics.
– Numerical values.
– May have multiple values.
• Visual Features
– Quantify visual characteristics of an image.
– Popular features.
• Colour, Texture, Shape
VC 13/14 - T18 - Visual Feature Extraction
Feature vector
• Feature Fi
• Feature Fi with N
values.
• Feature vector F with
M features.
• Naming conventions
for this module:
– Elements of a feature
vector are called
coefficients.
– Features may have
one or more
coefficients.
– Feature vectors may
have one or more
features.
ii fF
iNiii fffF ,...,, 21
MFFFF |...|| 21
VC 13/14 - T18 - Visual Feature Extraction
• Objective
• Directly reflect specific visual features.
– Colour
– Texture
– Shape
– Etc.
Low-level visual features
VC 13/14 - T18 - Visual Feature Extraction
Features & Decisions
Low
-Leve
l Featu
res
Mid
dle
-Leve
l Featu
res
Hig
h-L
eve
l Featu
res
Decision Decision
Various
Possible
Solutions
One
Solution
How do I
decide?
Starting point of most
classification tasks!
VC 13/14 - T18 - Visual Feature Extraction
How to quantify visual features?
– Many possibilities!
– We need a standard.
• MPEG-7 Standard
– Developed by the Moving Pictures Expert Group.
“is a standard for describing the multimedia content
data that supports some degree of interpretation of
the information meaning, which can be passed onto,
or accessed by, a device or a computer code” [MPEG-7 Overview (version 10),
ISO/IEC JTC1/SC29/WG11N6828]
VC 13/14 - T18 - Visual Feature Extraction
The MPEG-7 standard
• Provides a rich set of standardized tools to describe multimedia content.
– Computer annotation.
– Human annotation.
• Audiovisual Description Tools – Descriptors
– Descriptor Schemes
• Target functionality: – Efficient search, filtering and browsing of multimedia
content.
Feature
Vectors
VC 13/14 - T18 - Visual Feature Extraction
MPEG-7 Links
• MPEG website http://www.chiariglione.org/mpeg
• MPEG-7 Industry Forum website http://www.mpegif.com
• MPEG-7 Consortium website http://mpeg7.nist.gov
• MPEG-7 Overview (version 10) http://www.chiariglione.org/mpeg/standards/mpeg-7/mpeg-7.htm#E9E3
VC 13/14 - T18 - Visual Feature Extraction
Topic: Colour
• Feature Vectors
• Colour
• Texture
• Shape
VC 13/14 - T18 - Visual Feature Extraction
Gray-Level Histogram
• Intensity distribution (HSI).
• We can define the number of histogram bins.
• Histogram bins = Feature coefficients.
],...,[ 2550 ffF
VC 13/14 - T18 - Visual Feature Extraction
Colour Histogram
• We typically have three
histograms
Ex: RGB Colour space
- Red Histogram
- Green Histogram
- Blue Histogram
• How do we build a
feature vector?
– Concatenate vectors.
– Multi-dimensional
quantization of colour
space.
Red
Green
Blue
VC 13/14 - T18 - Visual Feature Extraction
RGB Histogram
• Simply concatenate vectors.
• Not very smart. (why?)
],...,[
],...,[
],...,[
2550
2550
2550
BBB
GGG
RRR
ffF
ffF
ffF
BGRRGB FFFF ||
VC 13/14 - T18 - Visual Feature Extraction
Combined Histogram
• Quantize multi-dimensional colour space.
• RGB
– Each coefficient is a small ‘cube’ inside the
RGB cube. ],...,[ 0 NffF
VC 13/14 - T18 - Visual Feature Extraction
HSI Histogram
• Quantize HSI space.
– Define number of bins
N.
– Feature vector
• Typically better for
object description
],...,[ 0 NHSI ffF
VC 13/14 - T18 - Visual Feature Extraction
MPEG-7 Dominant Colour
• Clusters colors into a small number of
representative colors (salient colors)
• F = { {ci, pi, vi}, s} • ci : Representative colors
• pi : Their percentages in the region
• vi : Color variances
• s : Spatial coherency
VC 13/14 - T18 - Visual Feature Extraction
MPEG-7 Scalable Colour
• HSI Histogram
• Typical quantization: 256 bins.
– 16 levels in H
– 4 levels in S
– 4 levels in I
• Very popular for CBIR (Content-Based
Image Retrieval).
],...,[ 2550 ffFSC
VC 13/14 - T18 - Visual Feature Extraction
[Sikora 2001]
VC 13/14 - T18 - Visual Feature Extraction
MPEG-7 Colour Layout
• Clusters the image into 64 (8x8) blocks
• Derives the average color of
each block (or using DCD)
• Applies (8x8)DCT and encoding
• Efficient for
– Sketch-based image retrieval
– Content Filtering using image indexing
VC 13/14 - T18 - Visual Feature Extraction
MPEG-7 Colour Structure
• Scanns the image by an
8x8 struct. element
• Counts the number of
blocks containing each
color
• Generates a color
histogram (HMMD/4CSQ
operating points)
8 x 8 structuringelement
COLOR BIN
C0
C1 +1
C2
C3 +1
C4
C5
C6
C7 +1
VC 13/14 - T18 - Visual Feature Extraction
Topic: Texture
• Feature Vectors
• Colour
• Texture
• Shape
VC 13/14 - T18 - Visual Feature Extraction
What is texture?
“Texture gives us information about the
spatial arrangement of the colours or
intensities in an image”.
[L. Shapiro]
VC 13/14 - T18 - Visual Feature Extraction
Two approaches to texture
• Structural approach
– Texture is a set of primitive texels in some
regular or repeated relationship.
– Good for regular, ‘man-made’ textures.
• Statistical approach
– Texture is a quantitative measure of the
arrangement of intensities in a region.
– More general and easier to compute.
VC 13/14 - T18 - Visual Feature Extraction
Statistical approaches
• Grey level of central pixels
• Average of grey levels in window
• Median
• Standard deviation of grey levels
• Difference of maximum and minimum grey levels
• Difference between average grey level in small and large windows
• Sobel feature
• Kirsch feature
• Derivative in x window
• Derivative in y window
• Diagonal derivatives
• Combine features How do I pick one??
VC 13/14 - T18 - Visual Feature Extraction
MPEG-7 Homogenous Texture
• Filters the image with a set of orientation
and scale sensitive filters.
• Computes mean and standard deviation of
response.
• 30 channels
– 6 in angular direction, 5 in radial direction.
30213021 ,...,,,,...,,,, dddeeeffF SCDCHT fDC, fSC are the mean intensity and the standard deviation of image texture), where
ex and dx are the logarithmically scaled texture energy and texture energy deviation coefficients.
VC 13/14 - T18 - Visual Feature Extraction
HT Channels
VC 13/14 - T18 - Visual Feature Extraction
MPEG-7 Local Edge Histogram
• Image divided into 4x4 sub-regions.
• Edge histogram computer for each sub-
region.
• Five bins:
– Vertical, horizontal, 45 diagonal, 135
diagonal, and isotropic.
• 80 total bins. ],...,[ 790 ffFLEH
VC 13/14 - T18 - Visual Feature Extraction
Topic: Shape
• Feature Vectors
• Colour
• Texture
• Shape
VC 13/14 - T18 - Visual Feature Extraction
Definitions
• Geometric definition
Two sets have the same shape if one can be
transformed into another by a combination of
translations, rotations and uniform scaling
operations.
VC 13/14 - T18 - Visual Feature Extraction
Shape and Segmentation
• Shape implies a segmentation step.
– Segmentation has multiple solutions (middle-
level feature).
– But the shape feature itself has a single
solution!
• How do we describe shapes?
– Chain-codes
– Statistical descriptors.
VC 13/14 - T18 - Visual Feature Extraction
Freeman Chain Code
• Chains represent the
borders of objects.
• Coding with chain
codes.
– Relative.
– Assume an initial
starting point for each
object.
• How do we build a
feature vector?
Freeman Chain
Code
Using a Freeman Chain Code and
considering the top-left pixel as
the starting point:
70663422
VC 13/14 - T18 - Visual Feature Extraction
MPEG-7 – Region-Based Shape
• Uses a set of separable ART (angular
radial transformation) functions.
• Classifies shape along various angular
and radial directions.
• Totals 35 coefficients.
],...,[ 340 ffFRBS
VC 13/14 - T18 - Visual Feature Extraction
ART Basis Functions
•Applicable to figures (a) – (e)
•Distinguishes (i) from (g) and (h)
•(j), (k), and (l) are similar
VC 13/14 - T18 - Visual Feature Extraction
MPEG-7 – Contour-Based
Descriptor • Finds curvature zero
crossing points of the shape’s contour (key points)
• Reduces the number of key points step by step, by applying Gaussian smoothing
• The position of key points are expressed relative to the length of the contour curve
VC 13/14 - T18 - Visual Feature Extraction
•Applicable to (a)
•Distinguishes differences in (b)
•Find similarities in (c) - (e)
Advantages:
• Captures the
shape very well
• Robust to the noise, scale,
and orientation
• It is fast and compact
VC 13/14 - T18 - Visual Feature Extraction
Comparison
• Blue: Similar shapes by Region-Based
• Yellow: Similar shapes by Contour-Based
VC 13/14 - T18 - Visual Feature Extraction
Resources
• L. Shapiro, Chapters 6 and 7
• T. Sikora, “MPEG-7 Visual Standard for
Content Description—An Overview”,
http://ieeexplore.ieee.org/iel5/76/20050/00
927422.pdf?arnumber=927422
top related